The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) because it processes untrusted source materials (transcripts and notes) and has file-system access via permitted tools.
Ingestion points: SKILL.md (Step 1) specifies reading full transcripts, notes, or briefs as the primary source material.
Boundary markers: The skill provides a structured format in SKILL.md (Step 9) that places transcripts within a fenced code block at the end of the document, which serves as a basic data boundary.
Capability inventory: The skill uses Read, Grep, Edit, and Write tools to modify files within the user's vault.
Sanitization: There are no specific instructions for the agent to sanitize content or explicitly ignore embedded commands within the ingested transcript data.

vault-scribe